The Planet Blog

After reading our last few posts, you know all about “the cloud” now, right? Well, yes and no. You know about the cloud in a general sense, but when it comes to current applications of the cloud, we need to drill down a little deeper.

The hosting industry is abuzz about cloud computing and cloud storage. Based on some completely fictional research*, 7 out of 10 hosting customers do not differentiate between the two. *The numbers may be fictional, but based on my experience working with customers, the sentiment is entirely true.

Virtualization and Abstraction

Before we get ahead of ourselves, let’s take a quick trip down memory lane … back to the days of virtualization. Why? Because the work put into developing virtualization has been a springboard for cloud technologies – specifically with regard to abstraction, the ability to present computational power and/or storage space without theoretical limits. Through abstraction, a single physical server can be divided into several distinct virtual servers, which function as independent physical servers that have their own dedicated resources.

Cloud computing and cloud storage take the principle of abstraction and tweak it. Instead of taking one physical server and creating several independent virtual servers, the development of the cloud takes multiple physical servers and creates virtual servers that freely move between physical machines as though they were all a single server. Naturally, that’s attractive to a hosting customer.

Cloud Computing

Cloud computing allows access to theoretically limitless computational resources. A user can scale from one Web server and one database server, to five Web servers and three database servers on the fly, with no upfront capital expenditure. Cloud computing essentially makes one large virtualized server that spans the entire available hardware infrastructure. Instead of having 20 servers with 4GHz of processor power each, the cloud shows 80GHz of processor power.

Cloud computing customers purchase a part of that cloud computing platform, and if no other customers are using resources on a given installation, that customer has the can use all 80GHz of process power one minute and scale back to almost nothing the next minute. The technology is in its infancy, but it’s helping to redefine the concept of a server: It’s not a black-and-white matter of physical resources anymore.

While the hardware abstraction is impressive, the greatest potential benefit of cloud computing is its use in Software as a Service (SaaS). It’s revolutionary to have an office application that scales from 10 users to 10,000 users and also available to anyone or any device with a network connection.

Cloud Storage

Cloud storage can be an amalgam of SaaS and HaaS (Hardware as a Service): Straightforward user interfaces combined with a solid hardware storage infrastructure. Because a cloud storage installation is dedicated to access, protection and serving data, the key component is hard disk space. Being able to pay for the space that meets your specific needs at a given time has significant advantages over a traditional solution like building out a storage area network in your local office. Your storage can be available to all of your satellite offices in London, Asmara and Santiago. Moreover, your company isn’t responsible for repairing file systems, replacing drives, or dealing with Nick Burns (your company’s computer guy). Brilliant!

We’ve launched a very successful cloud storage solution, and if you’re interested in seeing what the cloud can do for you, you can sign up for our Storage Cloud Test. On the cloud computing side, our team has been evaluating the most powerful, reliable and cost-efficient cloud computing solutions, and we plan on launching a computing platform as dominant as the storage platform in the near future.

Cloud computing and cloud storage have come a long way in the past year or two, and with the hosting industry’s focus on the development and enhancement of the platforms, the sky is the limit for the clouds.

Pun intended.

- Chris

Today, February 13, 2009, at exactly 5:31:30 p.m. CST, Unix users around the world will celebrate a momentous occasion: 1234567890 seconds will have passed since Unix time began at midnight UTC on January 1, 1970.

The relevance of this may be lost to some readers, but I assure you there is a very valid reason to celebrate such a seemingly inconsequential event. Many programs — including Operating Systems — measure time on this scale. And as you have all deduced, there will be only one time in the history of the 4byte epoch clock where we can see the sequence 1234567890. It’s party time!

I am holding out for 2147483647, when we will probably have to give up the system many of us hold dear … as it will let us down. In the year 2038, we will reach the maximum number possible in 32bits. As the time rolls toward that significant value, we’ll need to be prepared for it to run around and desert us. It’s like Y2K all over again … be VERY afraid, and brace for a possible Code Kermit.

If you want to count down the seconds until 1234567890 or just keep track of the current Unix time, you should check out this online Unix Epoch Clock.

-Phil

P.S. If you’re looking for a way to celebrate 1234567890, they just released the 1,234,567,890th Friday the 13th movie today.

You’ve probably been on pins and needles since you read our last blog post … anxiously anticipating the inside information of how we built a better-than-ideal-efficiency data center.

If you’re familiar with data center design, you’ve probably heard the terms “cold aisle” and “hot aisle.” A cold aisle is an aisle between racks of servers that sends cool air up through the floor (in the case of raised flooring data centers) and into the fronts of the servers. The air cools the server components and is exhausted through the back of the server as warm air, creating a hot aisle behind the server. Data centers are typically set up with racks of servers arranged front-to-front and back-to-back so that for every 2 rows of servers, only one cold aisle is needed. (If you’d like to see this setup in action, check out our popular Data Centric post.)

D6 Phase 3 uses those principles in a different way: hot and cold air are completely isolated.

Let’s take a look at how the phase is built to see what that looks like and why it is much more efficient.

In a raised-floor data center, your air conditioner blows air down under the floor into an air-tight compartment, and you insert vented floor tiles in the areas you want to cool. The large black unit in the image below is one of this phase’s air conditioners.

The metal posts you see on the ground are the braces used to create the grids upon which flooring tiles are installed.

The photo below is the bottom of a flooring tile. You’ll notice that there are no screws or bolts on the tile … it is simply laid on top of a grid of braces to make an air-tight seal. Each of the tiles is partially made out of concrete and is relatively heavy, so when a tile is installed, it’s not going anywhere unless you really want it to.

Below, you can see what the flooring grid looks like without floor tiles installed.

As we continue building the flooring, the data center appears to be taking shape … pretty standard process up to this point. But now, take a look at the air conditioning units below. Notice anything strange (aside from the fact that the covers aren’t installed)?

The return air plenum extends all the way to the ceiling of the room … now we’re getting somewhere.

I only retained a few things from my elementary school physics lessons, but one of them was that warm air rises and cold air sinks. This natural phenomenon is used in data center cooling: the air conditioners send the cold air down under the floor to cool the servers, when the servers send out warm air, the warm air rises to the top of the room, and the air conditioners pull the warm air from the top of the room to process and send back down as cold air. As Jeff mentioned in his interview, the higher the air conditioner pulls the air from, the warmer that air will be, and you don’t want to cool already cold air, so you should try and pull the hottest air in the data center.

Phase 3 goes a step further: it creates an airtight space above the ceiling tiles where all the warm air is exhausted and pulled in by the air conditioners.

So once we’ve got the floor tiles and the ceiling tiles installed, our fresh data center phase (without any server racks) looks like this:

This begs a big question: if the cold air is being sent through the floor to the servers and the warm air is being pulled from the ceiling, how do the servers pull the cold air and push the warm air without the heat being disseminated into the other areas of the data center?

Enter our new custom-made rack-mount cabinets.

The cabinet above is pulled out in one of our other data centers to demonstrate the design of the Phase 3 cabinets. The servers will pull cold air from the floor in front of the server but instead of exhausting the warm air out into an open aisle, the warm air will rise through a vent sealed to the ceiling (the sides, front and back of the cabinet are closed when the covers are installed).

So a completely installed server row looks like this:

One noticeable difference between this data center and the other data centers we’ve shown you in Houston and Dallas are the servers themselves: all rack-mount, no towers.

The rack-mount servers allow for a better power density throughout the data center, so we’ve got to make sure we can provide the power to all of our new servers in the event of a utility power outage.

We’ve got N+1 power redundancy, so for every phase, we have a dedicated backup generator, and for every data center we’ve got an extra backup generator in case any of the other generators fail. New phase = new generator. It’s pretty interesting to see the generator without its skin, right?

Now that power is accounted for, we can install our PDUs for each aisle and prepare to get servers up and running in the new phase.

A little networking and wiring, and D6 Phase 3 is ready for business! In the first picture below, you can see the orange tubes which are primary network drops into the phase … and yes, even those are sealed in the ceiling.

Head over to our The Planet’s Flickr page for a few more pictures of the new data center phase.

-Kevin

I think I’ve officially become a geek. I just got a few of the pictures from the build-out of our newest data center phase, and I’m almost ashamed to admit how excited I was to write this post as an opportunity to showcase them.

The Planet has six data centers between Houston and Dallas, and some have been built out in “phases” or semi-independent sections. This phased construction process allows us to invest our time and money more efficiently because we can create space as we need it, rather than building out an entire data center shell and having to heat/cool/power the excess space that may not actually house servers for months. Following a very welcome ramp up in demand for our dedicated servers and managed hosting, we recognized an impending need for more data center space, so our facilities team fired up their engines and began planning Phase 3 of our D6 data center … and that’s when the fun started.

Jeff Lowenberg, our vice president of facilities, spoke with WHIRtv about the progress we’ve made in improving our data center efficiency over the past year, and the new data center phase is a testament to the work Jeff and his team have done in researching and designing new data center space. With a few of these innovative ideas, we worked directly with our vendors to create custom DC solutions, and the results are beyond ideal … literally: the new phase’s coefficient of efficiency (the total power necessary to operate a data center divided by the power necessary to operate the servers alone) will be approximately 1.5, which is below (better than) the “ideal” ranking of 1.6. With a lower coefficient of efficiency, a greater percentage of our power is running servers rather than heating/cooling/lighting the data center space while maintaining a perfect operating environment for the servers.

Tune in tomorrow for a tour of the new phase’s construction where we’ll “show and tell” HOW this is possible.

-Kevin

P.S. You may have seen that we announced a new data center earlier this week – D7 – to be built out in Plano, Texas, just north of Dallas. We’re preleasing it and expect to have it ready in May, so you can expect to see a bit more info about that space on the blog as that facility takes shape.

In 1965, Intel co-founder Gordon Moore observed an interesting trend: “The complexity for minimum component costs has increased at a rate of roughly a factor of two per year … Certainly over the short term this rate can be expected to continue, if not to increase.”

Moore was initially noting the number of transistors that can be placed on an integrated circuit at a relatively constant minimal cost. Because that measure has proven so representative of the progress of our technological manufacturing abilities, “Moore’s Law” has become a cornerstone in discussions of pricing, capacity and speed of almost anything in the computer realm. You’ve probably heard the law used generically to refer to the constant improvements in technology: In two years, you can purchase twice as much capacity, speed, bandwidth or any other easily-measureable and relevant technology metric for the price you would pay today and for the current levels of production.

While I never questioned these assertions, I can’t say that I really investigated to see if Moore’s Law actually held true in the world of storage, especially with regard to the two key storage characteristics: capacity and throughput. Sure, prices for the same technology get lower over time – we all know that – but that’s just because no one wants the old stuff, right? Does Moore’s observation about the doubling of transistor density actually relate to hard drive capacities? What about throughput rates?

Once I started looking into historical storage-related statistics, it became clear that Moore’s Law doesn’t completely explain the evolution of storage technology. The primary driver for hard drive capacity – the disk’s areal density – has been increasing at 60 percent per year (or around 1.6x every two years), so that key metric of storage appears to correlate with Moore’s transistor observation, but drive speeds and seek times have not improved in a similar exponential manner.

This dichotomy may seem a little strange, but I think capacity limitations have been a more significant problem for the industry as a whole than throughput rates, so drive manufacturers have thrown more of their R&D budgets into improving that key characteristic first. The proven, constant increase in storage capacity reflects a focus on meeting user demand for that storage capacity, and if the incremental value of an additional gigabyte of storage decreases, I think we’ll see a similar improvement in throughput rates as manufacturers turn their focus to that other key storage characteristic. Moore’s observation focused on manufacturing with the single goal of more transistors on an integrated circuit, so we can’t really say Moore’s Law “doesn’t apply” to storage since hard drive manufacturers have several key measurements to improve at a given time.

Thanks to the trend Gordon Moore recognized 48 years ago, we were recently able to drop the prices on several of our backup products. EVault Backup is now priced at $1 per GB, down from $2, and Network Backup product is now priced at 50 cents per GB, down from $1, and they are both free for 90 days. If you’re interested in learning more about our data protection and backup options, check out my “What is Data Protection?” blog or leave me your questions in the comments section here.

-Rob

Over the course of the last several months, we’ve been working with Weaver & Tidwell, L.L.P., a highly-regarded certified public accounting firm out of Fort Worth, to complete an exhaustive Statement on Auditing Standards No. 70 (SAS 70) Type II audit. Developed by the American Institute of Certified Public Accountants (AICPA), the widely recognized auditing standard certifies that The Planet has been through a rigorous evaluation of its internal processes and controls through an independent third-party auditor.

Voluntarily undergoing an exhaustive audit by a third-party that takes months to complete.

A SAS 70 Type II audit is certainly a big-time undertaking. Some even think starting the process of a future review is worthy of a dedicated blog post … we just got it done.

In the process of the audit, we checked and evaluated the controls and processes for our network, customer provisioning systems, physical and environmental security, problem management and resolution through our customer portal, human resources department organization and administration, data center operations, and most importantly, our data centers themselves.

Daniel Golding, vice president and research director for Tier1 Research explains the significance of SAS 70 compliance in the context of the hosting industry:

Hosting providers that want to offer meaningful IT services to larger enterprises see SAS 70 as the means of both meeting Sarbanes-Oxley auditing requirements, while reassuring IT decision makers that their processes, facilities and staff are capable of providing true enterprise-grade services.

The Sarbanes-Oxley legislation consists of standards required of every public company and important to any company considering/anticipating an IPO. In searching for additional reference information on the significance of SAS 70 to SOX compliance, I came across a great resource: www.sas70.com. The site has a dedicated Sarbanes-Oxley page, where the significance of a Type 2 audit masterfully described:

Section 404 [of Sarbanes-Oxley] draws attention to the significant processes that feed and comprise the financial reporting process for an organization. In order for management to make its annual assessment on the effectiveness of its internal control, management is required to document and evaluate all controls that are deemed significant to the financial reporting processes. If the organization uses a service provider to process transactions, host data, or other signficant services, management may need to evaluate the design and test the operating effectiveness of the service organization’s controls.


Management will either need to conduct an evaluation of the service organization’s controls, or management may obtain a Type II SAS 70 service auditor’s report from the service organization, if a service auditor has been engaged, to gain an understanding of the service organization’s controls. The relevant audit guidance for SAS 70 already requires that a service auditor’s report contain information on the five components of internal control as it relates to the service organization.

The difference between a Type I audit and Type II audit is pretty significant: Both say “we have well-designed processes, controls and goals,” but the Type II audit must show that the controls and processes have been practiced and they were successful in achieving the initial goals. The proof is in the pudding.

What Does It Mean?

It’s clear that the successful completion of the SAS 70 Type II review is important to all of our customers. It reinforces our commitment to providing the best hosting experience in the industry. Our processes, practices, procedures and controls have been tested and have been proven successful in helping us achieve our operational goals.

-Kevin

Earth Day typically inspires widespread environmental introspection. How can we cut down on waste? Can we be more efficient? Are we actively pursuing “greener” operations? And how can we reduce our costs and be fiscally responsible?

Houston is recognized as the energy capital of the world, so it may be a surprise to learn that amidst that distinction The Planet does its part to reduce energy costs. In fact, we have been featured in several “green technology” articles over the past few months and acknowledged for our common sense approach. Ultimately, we look to save money, reduce consumption and improve data center efficiency. And in the coming weeks, we’ll announce an expanded program that takes us to the next level in increasing those efficiencies.

Tier1 is a leading research firm, and Martin Levy is the firm’s “green” analyst. In his report on The Planet, his headline was simple: “Down-to-earth solutions help improve efficiency at The Planet.”

Martin goes on to say the following:

Not a word about carbon offsets. Nobody planting trees. Nothing about Renewable Energy Credits (RECs). No recycling bins at the entrance to the datacenters. Instead, today’s announcement from The Planet was all about core datacenter efficiency. The company runs six datacenters and because of a focus on efficiency, it expects to save over one million dollars during 2008 … T1R is impressed. The Planet has shown that going green can be done the old-fashioned way. Make the technology work better and the company sees a positive ROI. That’s still good for the environment and even better for the bottom line!

Our facilities team is always on the lookout for new ways to reduce energy costs, since it’s one of our biggest expenses. Our vice president of facilities, Jeff Lowenberg, took an interesting challenge at the end of last year: Cut power costs by $1 million dollars in 2008, while we continue to grow and provision new servers in our six world-class data centers.

In his Sustainable IT blog, Ted Samson reported on a few of the initiatives aimed at improving our efficiency:

• Rearranging floor tiles to better manage cold airflow
• Installing seals and grommets in the ceilings, walls, and floors to reduce bypass airflow
• Installing blanking plates in server cabinets to direct airflow more efficiently
• Sealing power distribution units to reduce bypass airflow

Ted also explained the significance of those “minor” improvements:

Cool air was going to only where it was needed: the server intakes … Six months later, the company finds that its efforts have paid off substantially. Even though critical server loads increased by 5 percent, the facility’s overall cooling power needs dropped by 31 percent … The Planet also improved its “coefficient of efficiency,” an EPA- and Uptime Institute-recognized measurement of the total power necessary to operate a data center, divided by critical power, which represents the energy required to operate its computers. The company increased its rating to 1.7 – a near-ideal number – from its previous “good” ranking of 2.0.

Matt Stansberry at Search Data Center also spoke with Jeff about our progress and shared a few additional details in the quest to improve data center cooling:

Data center cooling is where most of infrastructure energy efficiency is lost. The fundamental rule in energy efficient cooling is to keep hot air and cold air separate … The Planet uses a method of extending the height of its computer room air conditioning (CRAC) units’ return-air plenums to optimize air cooling … By extending the plenums higher, it ensures that the CRAC units are not sucking in any cold air from the cold aisles, as it allows for the hottest air to be sucked into the units. In this scenario, the top of the plenums must be at least 2 feet from the ceiling.

To get an idea of what “plenums” are, you can visit Matt’s post or Heather Clancy’s recent article about The Planet at ZDNet’s GreenTech Pastures … and while you’re there, be sure to check out the post’s opening line.

To stay in the loop about what is being done in the “green tech” sphere, keep an eye on Ted Samson’s Sustainable IT blog, GreenerComputing, The Daily T1R from Tier1 Research, ZDNet’s GreenTech Pastures and Search Data Center.

And watch for more news from us.

-Yvonne